3. METHODS
The sample of 21 field plots had to be reduced to 18 plots due to problems and restrictions in the viewing geometry of the satellite system. The respective areas that cannot be viewed by TerraSAR-X shadow and layover were extracted
from the Geocoded Incidence Angle Mask GIM that is delivered with the EEC-product of TerraSAR-X. Backscattering from a target is influenced by the relative orientation of the illuminated ground cell and the sensor, as
well as by the distance in range between them. In order to minimize the impact of topography on the radar signal, the image pixel values were calibrated to normalized backscattering coefficients sigma-naught in dB using the equation
after [16]:
sin 10
log 10
loc dB
dB
θ β
σ +
=
1 and
10 log
10
2
DN k
s dB
= β
2 where
σ
dB
= backscattering coefficient in dB θ
loc
= local incidence angle from GIM β
dB
= radar brightness in dB K
s
= TerraSAR-X calibration factor DN
= digital number of TerraSAR-X EEC input image pixel. Against the background of the study, the heterogeneity in the calibrated image values was preserved and no filtering
was applied. This is also justified by the fact that the chosen processing format already partly accounts for speckle reduction see above. The acquired data were spatially reduced to the remaining field sites and subsequently first order
statistics mean, maximum, minimum, range, standard deviation, coefficient of variation were computed for each dataset within the regions of interest that are given by the extent of the field plots. In a second step textural parameters
second order statistics were computed based on the grey level co-occurrence matrix GLCM of the backscatter data after [17]. Due to the redundancy in texture measures, only contrast and entropy were chosen GLCM-shift 1x1 pixel.
The complete compilation of structure and texture measures provided the dataset of indirect variables for the estimation of geometric and canopy parameters. At total, nine indirect variables were analyzed together with 18 direct variables
from the field inventory. Correlation analysis was carried out and results were interpreted based on the correlation coefficient R and the level of significance p.
4. RESULTS
Kategori